Magnetic vortex echoes: application to the study of arrays of magnetic nanostructures F. Garcia1, J.P. Sinnecker2, E.R.P. Novais2, and A.P. Guimar˜aes2∗ 1Laborat´orio Nacional de Luz S´ıncrotron, 13083-970, Campinas, SP, Brazil and 2Centro Brasileiro de Pesquisas F´ısicas, 22290-180, Rio de Janeiro, RJ, Brazil (Dated: January 18, 2012) Weproposetheuseofthegyrotropicmotionofvortexcoresinnanomagnetstoproduceamagnetic echo, analogous to the spin echo in NMR. This echo occurs when an array of nanomagnets, e.g., nanodisks, is magnetized with an in-plane (xy) field, and after a time τ a field pulse inverts the 2 1 core magnetization; the echo is a peak in Mxy at t = 2τ. Its relaxation times depend on the inhomogeneity, on the interaction between the nanodots and on the Gilbert damping constant α. 0 Its feasibility is demonstrated using micromagnetic simulation. To illustrate an application of the 2 echoes, we have determined the inhomogeneity and measured the magnetic interaction in an array n of nanodisks separated by a distance d, findinga d−n dependence,with n≈4. a J PACSnumbers: 75.70.Kw,75.78.Cd,62.23.Eg,76.60.Lz 7 1 The interest in magnetic vortices and their properties ] and applications has grown steadily in the last years[1– l al 4]. Vortices have been observed, for example, in disks h and ellipses having sub-micron dimensions[5]. More re- - cently, the question of the intensity of the coupling be- s e tweenneighbordiskswithmagneticvortexstructureshas m attracted an increasing interest[6–9]. t. Atthevortexcorethemagnetizationpointsperpendic- a ularlytotheplane: thischaracterizesitspolarity,p=+1 m for the +z direction and p = 1 for z. The direc- - − − tionofthemomentsinthevortexdefinesthecirculation: d n c = 1 for clockwise (CW) direction, and +1 for CCW. − o If removed from the equilibrium position at the center c of the nanodisk by, for example, an in-plane field, and [ thenlefttorelax,avortexcorewillperformagyrotropic 1 motion, withangularfrequency ω,givenforthin disks[3] v by ω (20/9)γM β (β =h/R is the aspect ratio)[10]. G s 3 ≈ 5 We propose in this paper that, manipulating the dy- 5 namic properties of the vortex in an analogousway as it 3 is done in Nuclear Magnetic Resonance (NMR), a new 1. phenomenon results, the magnetic vortex echo (MVE), FIG. 1. (Color online) Diagram showing the formation of 0 similar to the spin echo observed in NMR[11]. This new magnetic vortex echoes; thedisks are described from a refer- 2 echomayprovideinformationonfundamentalproperties enceframe that turnswith theaverage translation frequency 1 of arrays of nanodisks, e.g., their inhomogeneity and in- ω0: a)diskswithin-planemagnetizationsMi alongthesame : direction (defined by the white arrows); b) after a time τ v teractions. Despite the fact that applications of vortices the disks on the left, center and right have turned with fre- i X necessarilyinvolvearrays,mostoftherecentpublications quencies, respectively, lower, higher and equal to ω0; c) the r deal with the analysis of individual nanodisks or arrays polarities of the vortex cores are reversed, and the ωi of the a with a few elements. Therefore, the possibility of char- vortex cores (and of theMi) changesign, and d) after a sec- acterizing large arrays is of much interest. In this paper ondintervalτ thecores(and Mi)areagain aligned, creating theecho. we have examined the motion of vortex cores in an ar- ray of nanometric disks under the influence of a pulsed magnetic field, using micromagnetic simulation. is the same: p =+1. This is notnecessaryfor ourargu- i Let us consider an array of nanodisks where the vor- ment, but, if required, the system can be prepared; see tex cores precess with a distribution of angular frequen- ref. [12] and the references therein. cies centered on ω , of width ∆ω, arising from any type Since the direction of rotation of the magnetic vortex 0 of inhomogeneity (see note [10]). We assume that the cores after removal of the in-plane field is defined exclu- frequencies vary continuously, and have a Gaussian dis- sively by p, all the cores will turn in the same direction; tribution P(ω) with mean square deviation∆ω. To sim- as the vortex core turns, the in-plane magnetization of plifywecanassumethatthe polarizationofeveryvortex the nanodot also turns. 2 one can derive the total in-plane magnetization: 1 = 10 nm, = 30 ns, = 0 0 ∞ eh−21(ω∆−ωω20)2i -1 a My(t)=My(0) cos(ωt)dω (1) Z−∞ ∆ω√2π = 20 nm, = 10 ns, = 40 ns, = 0.001 1 m) an integral that is[16] the Fourier transform of the func- -14 0A/-01 b tion P(ω); using T2∗ =1/∆ω: 1 M ( 1 = 20 nm, = 20 ns, = 0 1 t2 M (t)=M (0) exp cos(ω t) (2) 0 y y (cid:18)−2T2∗2(cid:19) 0 -1 c This result shows that the total magnetization tends to 1 = 10 nm, = 20 ns, = 0.005 zero, as the different contributions to My(t) get gradu- ally out of phase. This decay is analogous to the free 0 induction decay (FID) in NMR; its characteristictime is -1 d ∗ T =1/∆ω. 2 0 10 20 30 40 50 60 70 80 90 After a time τ, the angle rotated by each vortex core t (ns) will be ωτ; if at t=τ we invertthe polarities of the vor- FIG.2. (Coloronline)Micromagneticsimulationofmagnetic ticesinthe array,usinganappropriatepulse,themotion vortex echoes, for 100 nanodisks, with d = infinity, and a) ofthecoreswillchangedirection(i.e.,ω ω),andone σ = 10nm, τ = 30ns, α = 0; b) σ = 20nm, τ = 10ns →− obtains: and τ = 40ns (two pulses), and α = 0.001; c) σ = 20nm, τ = 20ns, α = 0; d) σ = 10nm, τ = 20ns, α = 0.005. The ∞ −1(ω−ω0)2 inversion pulses (Bz =−300mT) are also shown (in red). M (t τ)=M (0) e 2 ∆ω2 cos[ω(τ t)]dω (3) y y − Z−∞ ∆ω√2π − The magnetization at a time t>τ is then: 0.10 My(t)=My(0) e[−21(t−T22∗τ2)2]e(−tT−2τ)cos(ω0t) (4) 0.08 ThismeansthatthemagnetizationcomponentM (t)in- y creases for τ < t < 2τ, reaching a maximum at a time 0.06 -1 s) t=2τ: this maximumis the magnetic vortex echo, anal- n T (20.04 ogous to the spin echo observed in magnetic resonance, 1/ which has important applications in NMR, including in Magnetic Resonance Imaging (MRI)[14, 15] (Fig. 1). In 0.02 thecaseoftheNMRspinecho,themaximumarisesfrom the refocusing of the in-plane components of the nuclear magnetization. 0.00 In Eq. 4 we have included a relaxation term contain- 0.000 0.001 0.002 0.003 0.004 0.005 ing the time constant T - also occurring in NMR -, to 2 Damping constant account for a possible decay of the echo amplitude with time; its justification will be given below. FIG. 3. (Color online) Variation of the inverse of the re- laxation times T2 (diamonds) obtained by fitting the curves Inthearrayofnanodisks,therewillbeinprincipletwo of echo intensity versus time interval τ to M0exp(−τ/T2), contributionstothedefocusingofthemagnetization,i.e., as a function of α, for simulations made with D = 250nm, two mechanisms for the loss of in-plane magnetization σ =10nm, separation infinite; the continuous line is a linear memory: 1) the spread in values of β and H (see note least squares fit. [10]), producing an angular frequency broadening term ∆ω, and 2) irreversible processes that are characterized ∗ by a relaxation time T : thus 1/T =∆ω+1/T . 2 2 2 Let us assume that all the vortex cores have been dis- The second contribution is the homogeneous term placedfromtheirequilibriumpositionsalongthepositive whose inverse, T , is the magnetic vortex transverse re- 2 x axis[13], by a field B. The total in-plane magnetiza- laxationtime,analogoustothespintransverserelaxation tion (that is perpendicular to the displacement of the time (or spin-spin relaxation time) T in magnetic reso- 2 core) will point along the y axis, therefore forming an nance. The irreversible processes include a) the interac- angle θ = 0 at t = 0. Using the approach employed in tion between the disks, which amounts to random mag- 0 the descriptionof magnetic resonance(e.g., see [14, 15]), netic fields that will increase or decrease ω of a given 3 ′ ′ disk,producingafrequencyspreadofwidth∆ω =1/T , we have introduced a Gaussian distribution of diame- 2 and b) the loss in magnetization (of rate 1/T ) arising ters, centered on 250nm and mean square deviation σ; α fromthe energydissipationrelatedto the Gilbert damp- σ = 10nm corresponds to ∆ω 1.6 108s−1. The ≈ × ing constant α. Identifying T to the NMR longitudinal disks were placed at random on the square lattice. The α ′ relaxation time T , one has [14]: 1/T =1/T +1/2T . initial state of the disks (p = +1 and c = 1) was pre- 1 ∗2 2 α − Therefore the relaxation rate 1/T is given by: pared by applying an in-plane field of 25mT; the po- 2 larity was inverted with a Gaussian pulse of amplitude 1 1 1 1 =∆ω+ =∆ω+ + (5) B = 300mT, with width 100ps. The results for the T∗ T T′ 2T z − 2 2 2 α case d= were simulations made on the disks one at a ∞ 1/T∗ isthereforethe totalrelaxationrateofthe in-plane time, adding the individual magnetic moments µi(t). 2 magnetization, composed of a) ∆ω, the inhomogeneity We have successfully demonstrated the occurrence of term,andb)1/T ,thesumofalltheothercontributions, the magnetic vortex phenomenon, and have shown its 2 containing1/T′,duetotheinteractionbetweenthedisks, potential as a characterization technique. The simula- 2 and 1/T , the rate of energy decay. The vortex cores tions have confirmed the occurrence of the echoes at the α will reach the equilibrium position at r = 0 after a time expected times (t = 2τ). For different values of σ, the ∗ t Tα, therefore there will be no echo for 2τ Tα. T2 time, and consequently the duration of the FID and ∼The vortex echo maximum at t = 2τ, from≫Eq. 4, is the width ofthe echoaremodified (Fig. 2a,2c); increas- M (2τ) exp( τ/T ); one should therefore note that ing α results in a faster decay of the echo intensity as a y 2 the maxi∝mum m−agnetization recovered at a time 2τ de- function of time (Fig. 2a, 2d). We have also obtained creasesexponentiallywith T , i.e., this maximum is only multiple echoes, by exciting the system with two pulses 2 affectedby the homogeneouspartofthe totaldecayrate (Fig. 2b)[18]. given by Eq. 5. In other words, the vortex echo cancels Fig. 3showsthedependenceofT2 onαforσ =10nm; the loss inM due to the inhomogeneity∆ω, butit does essentially the same result is obtained for σ = 20nm, y not cancel the decrease in My due to the interaction be- sinceT2 doesnotdependon∆ω (Eq. 5). Takingalinear tween the nanodisks (the homogeneous relaxation term approximation, 1/T2 = Aα, and since for d =infinity 1/T2′), or due to the energy dissipation (term 1/2Tα). there is no interaction between the disks, 1/T2 =1/2Tα, Note also that if one attempted to estimate the inho- and therefore: mogeneityofanarrayofnanodotsusinganothermethod, 1 =2Aα (6) for example, measuring the linewidth of a FMR spec- T α trum, one would have the contribution of this inhomo- From the least squares fit (Fig. 3), A = 1.6 1010s−1. geneitytogetherwiththe othertermsthatappearinEq. × Thisrelationcanbeusedtodetermineexperimentallyα, 5, arising from interaction between the dots and from measuring T with vortex echoes, for an array of well- the damping. On the other hand, measuring the vor- 2 separated disks. tex echo it would be possible to separate the intrinsic Recently some workers have analyzed the important inhomogeneity from these contributions, since T can be 2 problemoftheinteractionbetweendisksexhibitingmag- measured separately, independently of the term ∆ω. T 2 netic vortices,obtaining thatitvarieswith a d−n depen- can be measured by determining the decay of the echo dence: Vogel and co-workers [6], using FMR, obtained amplitude for different values of the interval τ. for a 4 300 array a dependence of the form d−6, the The Fourier transform of either the vortex free induc- × same found by Sugimoto et al. [8] using a pair of disks tiondecayorthetimedependenceoftheechoM (t)gives y excitedwith rfcurrent. Junget al. [7]studying a pairof the distribution of gyrotropic frequencies P(ω). nanodisks with time-resolved X-ray spectroscopy, found For the experimental study of vortex echoes, the se- n = 3.91 0.07 and Sukhostavets et al. [9], also for quence of preparation (at t=0) and inversion fields (at ± a pair of disks, in this case studied by micromagnetic t=τ) should of course be repeated periodically, with a simulation, obtained n = 3.2 and 3.7 for the x and y period T T . As in pulse NMR, this will produce α ≫ interaction terms, respectively. echoes on every cycle, improving the S/N ratio of the ∗ As a first approximation one can derive the depen- measuredsignals. Also note that the time T canbe ob- 2 ∗ denceofthe contributionto 1/T relatedtothe distance tained either fromthe initial decay (FID, Eq. 2) or from 2 between the disks as: the echo (Eq. 4), but T can only be obtained from the 2 MVE. In order to demonstrate the MVE, we have performed T∗ =B+Cd−n (7) micromagnetic simulations of an assembly of 100 mag- 2 netic nanodisks employing the OOMMF code[17]. The From our simulations, and using Eq. 7 we found, from simulated system was a square array of 10 10 disks, the best fit, that this interactionvaries asd−n, with n= × thickness 20nm, with distance d from center to center. 3.9 0.1, in a good agreement with [7] and reasonable ± In order to simulate the inhomogeneity of the system, agreement with Sukhostavets et al.[9]. 4 2 m) systems. 6.5 14 ment x 10(A 000...012 ccoielTslahCbeoNraPautqtih,oonCr;AswPweEoauSrle,dFaAlliskoPeEintRodJetb,htFaeAndkPtoEGSt.hMPe..BB.raFziioliranfoargtehne- 6.0 o 5.5 gnetic m--00..21 ns) Ma 0 10 20 30 40 50 60 70 ∗ Author to whom correspondence should be T* (25.0 Time (ns) addressed: [email protected] [1] A. P. Guimara˜es, Principles of Nanomagnetism 4.5 (Springer, Berlin, 2009) [2] C. L. Chien, F. Q. Zhu, and J.-G. Zhu, Physics Today 4.0 60, 40 (2007) [3] K. Y. Guslienko, K.-S. Lee, and S.-K. Kim, Phys. Rev. 3.5 Lett. 100, 027203 (2008) [4] F. Garcia, H.Westfahl, J. Schoenmaker, E. J. Carvalho, 0 1 2 3 4 5 6 A. D. Santos, M. Pojar, A. C. Seabra, R. Belkhou, -4 -11 -4 d (10 nm ) A. Bendounan, E. R. P. Novais, and A. P. Guimara˜es, Appl. Phys.Lett. 97, 022501 (2010) FIG. 4. (Color online) Variation of the relaxation times T2∗ versus d−4 for an array of 10×10 magnetic nanodisks [5] EM.aRrt.inPs.,FN.oGvaairsc,iPa,.aLnadndAe.rPos.,GAu.imGa.rSa˜e.sB,Jar.bAopsap,l.MP.hyDs.. with a distribution of diameters centered on D = 250nm 110, 053917 (2011) (σ = 10nm), damping constant α = 0.001 and separation [6] A. Vogel, A. Drews, T. Kamionka, M. Bolte, and d between their centers; the continuous line is a linear least G. Meier, Phys.Rev.Lett. 105, 037201 (2010) squares fit. The inset shows a vortex echo simulation for the [7] H. Jung, K.-S. Lee, D.-E. Jeong, Y.-S. Choi, Y.-S. array, with d=500nm,τ =30ns, α=0.001. Yu, D.-S. Han, A. Vogel, L. Bocklage, G. Meier, M.- Y. Im, P. Fischer, and S.-K. Kim, Sci. Rep. 59, 1 (2011/08/10/online) ′ Determining 1/T has allowed us to obtain the inten- 2 [8] S. Sugimoto, Y. Fukuma, S. Kasai, T. Kimura, A. Bar- sity of the interaction between the disks as a function man,andY.Otani,Phys.Rev.Lett.106,197203 (2011) of separation d between them. Substituting Eq. 6 and [9] O.V.Sukhostavets,J.M.Gonzalez,andK.Y.Guslienko, Eq. 7 in Eq. 5, we can obtain the expression for the Appl. Phys.Express 4, 065003 (2011) interaction as a function of d: [10] The sources of inhomogeneity are the spread in radii, in thicknessorthepresenceofdefects.Anexternalperpen- 1 1 dn dicularfieldHaddsacontributiontoω[19],ω=ωG+ωH, T′ = B+Cd−n −Aα−∆ω ≈ C −Aα−∆ω; (8) with ωH =ω0 p(H/Hs), where p is the polarity and Hs 2 | | the field that saturates the nanodisk magnetization. A distribution ∆H is another source of thespread ∆ω (approximationvalidfor d small). InFig. 4 we showthe [11] E. L. Hahn,Phys.Rev.80, 580 (1950) resultsofthesimulationswithσ =10nmandα=0.001. [12] R. Antos, M. Urbanek, and Y. Otani, J. Phys.: Conf. Assuming n = 4 and making a linear squares fit, we Series 200, 042002 (2010) obtained B =6.15 10−9s, C = 4.03 10−35s m4. [13] Ifthediskshavedifferentcirculations(c=±1)thecores × − × A new phenomenon, the magnetic vortex echo, anal- willbedisplacedinoppositedirections,buttheeffectwill ogous to the NMR spin echo, is proposed and demon- be the same, since the Mi will all point along the same direction. strated here through micromagnetic simulation. Appli- [14] C. P. Slichter, Principles of Magnetic Resonance, 3. ed. cationsofthemagneticvortexechoincludesthemeasure- (Springer, Berlin, 1990) ment of the inhomogeneity, such as, distribution of di- [15] A.P.Guimara˜es,MagnetismandMagneticResonancein mensions,aspectratios,defects, andperpendicularmag- Solids (John Wiley & Sons, New York,1998) netic fields and so on, in a planar array of nanodisks or [16] T. Butz, Fourier Transformation for Pedestrians ellipses; it may be used to study arrays of nanowires or (Springer, Berlin, 2006) nanopillars containing thin layers of magnetic material. [17] Available from http://math.nist.gov/oommf/ [18] These echoes, however, are not equivalent to the stimu- These properties cannot be obtained directly, for exam- lated echoes observed in NMR with two 90o pulses[11] ple, from the linewidth of FMR absorption. [19] G. de Loubens, A. Riegler, B. Pigeau, F. Lochner, The MVE is a tool that can be used to evaluate the F. Boust, K. Y. Guslienko, H. Hurdequint, L. W. interactionbetweentheelementsofalargearrayofnano- Molenkamp, G. Schmidt, A. N. Slavin, V. S. Tiberke- magnets with vortex ground states. It can also be used vich,N.Vukadinovic,andO.Klein,Phys.Rev.Lett.102, to determine the Gilbert damping constant α in these 177602 (2009)